The Global Database of Data Center Industry Expertise
ToE NVMeoF TCP Performance Line Boost Performance Reduce Costs.
Yes, you read that correct; leverage TCP offload Engines (TOE) to boost the performance of TCP-based NVMeoF (e.g., NVMe over Fabrics) while reducing costs. Keep in mind that there is a difference between cutting costs (something that causes or moves problems and complexities elsewhere) and reducing and removing costs (e.g., finding, fixing, removing complexities).
Reducing or cutting costs can be easy by simply removing items for lower-priced items and introducing performance bottlenecks or some other compromise. Likewise, boosting performance can be addressed by throwing (deploying) more hardware (and or software) at the problem resulting in higher costs or some other compromise.
On the other hand, as mentioned above, finding, fixing, removing the complexity and overhead results in cost savings while doing the same work or enabling more work done via the same costs, maximizing hardware, software, and network costs. In other words, a better return on investment (ROI) and a lower total cost of ownership (TCO).
With the continued shift towards software-defined data centers, software-defined data infrastructures, software-defined storage, software-defined networking, and software-defined everything, those all need something in common, and that is hardware-based compute processing.
In the case of software-defined storage, including standalone, shared fabric or networked-based, converged infrastructure (CI) or hyper-converged infrastructure (HCI) deployment models, there is the need for CPU compute, memory, and I/O, in addition to storage devices. This means that the software to create, manage, and perform storage tasks needs to run on a server's CPU, along with I/O networking software stacks.
However, it should be evident that sometimes the obvious needs to be restarted, which is that software-defined anything requires hardware somewhere in the solution stack. Likewise, depending on how the software is implemented, it may require more hardware resources, including server compute, memory, I/O, and network and storage capabilities.
Keep in mind that networking stacks, including upper and lower-level protocols and interfaces, leverage software to implement their functionality. Therefore, the value proposition of using standard networks such as Ethernet and TCP is the ability to leverage lower-cost network interface cards (or chips), also known as NICs combined with server-based software stacks.
On the one hand, costs can be reduced by using less expensive NICs and using the generally available server CPU compute capabilities to run the TCP and other networking stack software. On systems with a lower application or other software performance demands, this can work out ok. However, for workloads and systems using software-defined storage and other applications that compete for server resources (CPU, memory, I/O), this can result in performance bottlenecks and problems.
There is a classic saying that the best I/O is the one that you do not have to do. Likewise, the second-best I/O is the one with the most negligible overhead (and cost) as well as best performance. Another saying is that many application, database, server, and storage I/O problems are actually due to CPU bottlenecks. Fast storage devices need fast applications on fast servers with fast networks. This means finding and removing blockages, including offloading server CPU from performing network I/O processing using TOEs.
Wait a minute, isn't the value proposition of using software-defined storage or networking to use low-cost general-purpose servers instead of more expensive hardware devices? With some caveats, Yup understands how much server CPU us being used to run the software-defined storage and software stacks and handle upper-level functionality. To support higher performance or larger workloads can be putting in more extensive (scale-up) and more (scale-out) servers and their increased connectivity and management overhead.
This is where the TOEs come into play by leveraging the best of both worlds to run software-defined storage (and networking) stacks, and other software and applications on general-purpose compute servers. The benefit is the TCP network I/O processing gets offloaded from the server CPU to the TOE, thereby freeing up the server CPU to do more work or enabling a smaller, lower-cost CPU to be used.
After all, many servers, storage, and I/O networking problems are often server CPU problems. An example of this is running the TCP networking software stack using CPU cycles on a host server that competes with the other software and applications. In addition, as an application does more I/O, for example, issuing reads and write requests to network and fabric-based storage, the server's CPUs are also becoming busier with more overhead of running the lower-layer TCP and networking stack.
The result is server resources (CPU, memory) are running at higher utilization; however, there is more overhead. Higher resource utilization with low or no overhead, low latency, and high productivity are good things resulting in lower cost per work done. On the other hand, high CPU utilization, server operating system or kernel mode overhead, poor latency, and low productivity are not good things resulting in host per work done.
This means there is a loss of productivity as more time is spent waiting, and the cost to do a unit of work, for example, an I/O or transaction, increases (there is more overhead). Thus, offload engines (chips, cards, adapters) come into play to shift some software processing from the server CPU to a specialized processor. The result is lower server CPU overhead leaving more server resources for the main application or software-defined storage (and networking) while boosting performance and lowering overall costs.
Offload engines are not new, they have been around for a while, and in some cases, more common than some realize going by different names. For example, graphical Processing Units (GPUs) are used for offloading graphic and compute-intensive tasks to special chips and adapter cards. Other examples of offload processors include networks such as TCP Offload Engine (TOE), compression, and storage processing, among others.
The basic premise of offload engines is to move or shift processing of specific functions from having their software running on a general-purpose server CPU to a specialized processor (ASIC, FPGA, adapter, or mezzanine card). By moving the processing of functions to the offload or unique processing device, performance can be boosted while freeing up a server's primary processor (CPU) to do other useful (and productive) work.
There is a cost associated with leveraging offloads and specialized processors; however, the business benefit should be offset by reducing primary server compute expenses or doing more work with available resources and driving network bandwidth line rates performance. The above should result in a net TCO reduction and boost your ROI for a given system or bill of material, including hardware, software, networking, and management.
Ethernet network TOEs became popular in the industry back in the early 2000s, focusing on networked storage and storage networks that relied on TCP (e.g., iSCSI).
Fast forward to today, and there is continued use of networked (ok, fabric) storage over various interfaces, including Ethernet supporting different protocols. One of those protocols is NVMe in NVMe over Fabrics (NVMeoF) using TCP and underlying Ethernet-based networks for accessing fast Solid State Devices (SSDs).
An example of server storage I/O network TOEs, including those to support NVMeoF, are those from Chelsio Communications, such as the T6 25/100Gb devices. Chelsio announced today server storage I/O benchmark proof points for TCP based NVMe over Fabric (NVMeoF) TOE accelerated performance. StorageIO had the opportunity to look at the performance-boosting ability and CPU savings benefit of the Chelsio T6 prior to todays announcement.
After reviewing and validating the Chelsio proof points, test methodology, and results, it is clear that the T6 TOE enabled solution boosts server storage I/O performance while reducing host server CPU usage. The Chelsio T6 solution combined with Storage Performance Development Kit (SPDK) software, provides local-like performance of network fabric distributed NVMe (using TCP based NVMeoF) attached SSD storage while reducing host server CPU consumption.
“Boosting application performance, efficiency, and effectiveness of server CPUs are key priorities for legacy and software defined datacenter environments,” said Greg Schulz, Sr. Analyst Server Storage. “The Chelsio NVMe over Fabrics 100GbE NVMe/TCP (TOE) demonstration provides solid proof of how high-performance NVMe SSDs can help datacenters boost performance and productivity, while getting the best return on investment of datacenter infrastructure assets, not to mention optimize cost-of-ownership at the same time. It’s like getting a three for one bonus value from your server CPUs, your network, and your application perform better, now that’s a trifecta!”
You can read more about the technical and business benefits of the Chelsio T6 TOE enabled solution along with associated proof points (benchmarks) in the PDF white paper found here and their Press Release here. Note that the best measure, benchmark, proof point, or test is your application and workload, so contact Chelsio to arrange an evaluation of the T6 using your workload, software, and platform.
Learn more about TOE, server, compute, GPU, ASIC, FPGA, storage, I/O networking, TCP, data infrastructure and software defined and related topics, trends, techniques, tools via the following links:
Chelsio Communications T6 Performance Press Release (PDF) Chelsio Communications T6 TOE White Paper (PDF) Application Data Value Characteristics Everything Is Not the Same PACE your Infrastructure decision-making, it's about application requirements Data Infrastructure Server Storage I/O Tradecraft Trends Data Infrastructure Overview, Its What’s Inside of Data Centers Data Infrastructure Management (Insight and Strategies)Additional learning experiences along with common questions (and answers), as well as tips can be found in Software Defined Data Infrastructure Essentials book.
The large superscalar web services and other large environments leverage offload engines and specialized processing technologies (chips, ASICs, FPGAs, GPUs, adapters) to boost performance while reducing server compute costs or getting more value out of a given server platform. If it works for the large superscalars, it can also work for your environment or your software-defined platform.
The benefits are reducing the number and cost of your software-defined platform bill of materials (BoM). Another benefit is to free up server CPU cycles to run your storage or network or other software to get more performance and work done. Yet another benefit is the ability to further stretch your software license investments, getting more work done per software license unit.
Have a look at the Chelsio Communications T6 line of TOE for NVMeoF and other workloads to boost performance, reduce CPU usage and lower costs. See for yourself The TOE NVMeoF TCP Performance Line Boost Performance Reduce Costs benefit.
Ok, nuff said, for now. Cheers GSGreg Schulz - Microsoft MVP Cloud and Data Center Management, previous 10 time VMware vExpert. Author of Software Defined Data Infrastructure Essentials (CRC Press), Data Infrastructure Management (CRC Press), as well as Cloud and Virtual Data Storage Networking (CRC Press), The Green and Virtual Data Center (CRC Press), Resilient Storage Networks (Elsevier) and twitter @storageio. Courteous comments are welcome for consideration. First published on https://storageioblog.com any reproduction in whole, in part, with changes to content, without source attribution under title or without permission is forbidden.
All Comments, (C) and (TM) belong to their owners/posters, Other content (C) Copyright 2006-2021 Server StorageIO and UnlimitedIO. All Rights Reserved. StorageIO is a registered Trade Mark (TM) of Server StorageIO.
The following is a new Industry Trends Perspective White Paper Report titled ROI From Use Of Global Control Plane For Expanding VDI Environments.
This new StorageIO report looks at ROI From Use Of Global Control Plane For Expanding VDI environments. Using a Pro-Forma analysis this report provides a financial economic model comparison with Return on Investment (ROI) cost savings analysis for managing cloud based virtual desktop infrastructures (VDI) environments.
IT data infrastructure resource (servers, storage, I/O network, hardware, software, services) decision-making involves evaluating and comparing technical attributes (speeds, feeds, features) of a solution or service. Another aspect of data infrastructure resource decision-making involves assessing how a solution or service will support and enable a given application workload, along with associated management costs from a Performance, Availability, Capacity, and Economic (PACE) perspective.
Keep in mind that all application workloads have some amount of PACE resource requirements that may be high, low or various permutations, along with associated management costs. Performance, Availability (including data protection along with security) as well as Capacity are addressed via technical speeds, feeds, functionality along with workload suitability analysis.
Management costs are a function of initial and recurring tasks to support a given function or service such as VDI. The cost of management includes staff salary, along with amount of time needed to perform various tasks. The E in PACE resource decision-making is about the Economic analysis of various costs associated with different solution approaches.
The above image is an example from the White Paper Report titled ROI From Use Of Global Control Plane For Expanding VDI Environments.
In the example shown above, 36 month OpEx cost (and time) savings are shown using traditional cloud based VDI management tools, technologies and techniques vs. a modern cloud platform integrated global control plane solution. Leveraging a cloud platform integrated global control plane solution such as NetApp VDS among others, management costs can be reduced for initial and recurring tasks from $2,587,394 to $968,041 for 1,001 users.
In addition to the cost savings shown above, note the reduction in management hours of 21,653 over 36 months which could be used for doing other work, or reducing your OpEx spend. Of course your savings will vary based on what tasks, time per task, admin cost among other considerations.
The shift from Capital Expenditures (e.g. CapEx) IT data infrastructure spending to Operational Expenditures (e.g. OpEx) focus particular with IT clouds has resulted in increased OpEx budget demands. Increased spending is more than simply moving IT spend from the CapEx to OpEx columns in budgets. OpEx increases are a cumulation of increased cloud services and data infrastructure spend, along with management (initial and recurring) costs.
The good news is that there are OpEx opportunities to reduce, or, stretch your IT budget to do more while boosting productivity, performance, and effectiveness without compromise. By looking at how to use new technologies in new ways, including leverage cloud platform integrated global control planes for management of VDI (and other functions), initial and recurring OpEx management costs can be reduced.
Read more in this Server StorageIO Industry Trends Report here.
Learn more about ROI From Use Of Global Control Plane For Expanding VDI Environments, Clouds and Data Infrastructure related trends, tools, technologies and topics via the following links:
Application Data Value Characteristics Everything Is Not the Same PACE your Infrastructure decision-making, it's about application requirements Cloud conversations: confidence, certainty, and confidentiality Industry adoption vs. industry deployment, is there a difference? Ten tips to reduce your cloud compute storage costs Don't Stop Learning Expand Your Skills Experiences Everyday Data Infrastructure Server Storage I/O Tradecraft Trends Data Infrastructure Overview, Its What’s Inside of Data Centers Data Infrastructure Management (Insight and Strategies) Data Protection Diaries (Archive, Backup, BC, BR, DR, HA, Security) NetApp VDS with Global Control Plane Cloud VDI ManagementAdditional learning experiences along with common questions (and answers), as well as tips can be found in Software Defined Data Infrastructure Essentials book.
In addition, looking at your IT data infrastructure cloud spend can also help you to boost the effectiveness, productivity and return on investment while reducing your OpEx spend, or doing more with it. Leveraging financial pro-forma analysis as a tool in conjunction with your technology feature function, speeds, feeds comparisons enables informed decision making.
When comparing and making data infrastructure resource decisions, consider the application workload PACE characteristics. Shift or expand your focus from simply looking at costs from a efficiency utilization perspective to also include performance, productivity, and effectiveness of your IT OpEx spending.
Keep in mind that PACE means Performance (productivity), Availability (data protection), Capacity and Economics. This includes making decisions from a technical feature, functionality (speeds and feeds) capacity as well as how the solution supports your application workload. Leverage resources including tools to perform analysis including ROI From Use Of Global Control Plane For Expanding VDI Environments approaches.
Ok, nuff said, for now. Cheers GSGreg Schulz - Microsoft MVP Cloud and Data Center Management, previous 10 time VMware vExpert. Author of Software Defined Data Infrastructure Essentials (CRC Press), Data Infrastructure Management (CRC Press), as well as Cloud and Virtual Data Storage Networking (CRC Press), The Green and Virtual Data Center (CRC Press), Resilient Storage Networks (Elsevier) and twitter @storageio. Courteous comments are welcome for consideration. First published on https://storageioblog.com any reproduction in whole, in part, with changes to content, without source attribution under title or without permission is forbidden.
All Comments, (C) and (TM) belong to their owners/posters, Other content (C) Copyright 2006-2021 Server StorageIO and UnlimitedIO. All Rights Reserved. StorageIO is a registered Trade Mark (TM) of Server StorageIO.
2020 Holidays Seasons Greetings from StorageIO. Here's a video link from our compansion sister site Pictures Over Stillwater featuring drone based aerial footage from downtown Stillwater Minnesota on the St Croix river and inaugural lighting of the nearly 100 year old historic lift bridge, along with Stillwater Lights.
Downtown Stillwater Lights Night Bridge LightingIn addition to being an FAA part 107 certified (e.g. licensed) commercial drone operator, Im also authorized for night flying in the U.S with a daylight waiver (107.29). Thus the following night drone images were captured using my daylight waiver. The first 30 seconds or so of the following video was shot in 4K, the remainder in 6K rendered down for subsequent posting.
The following sunset aerial view is from downtown St Paul over looking Rice Park across from the Landmark Center with the Minnesota State Capital in the background (top left). Note that this location is a "0" zero altitude drone no fly zone due to proximity to downtown St Paul Airport, however I have an FAA airspace authorization and applicable state licenses to fly in this area.
You might wonder what do the above images other than having a holiday theme to them have to do with StorageIO?
Simple, all of the primary big data captured by the IoT devices, sensors, cameras and associated telemetry and metadata need to be moved, stored, processed, and protected somewhere which is one of the focuses of StorageIO. Meanwhile Pictures Over Stillwater has as one of its focuses, capturing and generating data including 6K video, 360 still and video images, VR, photogrammetry (2D/3D ortho models) among other rich content.
Learn and view more via the following links:
Additional learning experiences along with common questions (and answers), as well as tips can be found in Software Defined Data Infrastructure Essentials book.
2020 Happy Holidays Seasons Greetings from StorageIO
Ok, nuff said, for now.
Cheers GS
Greg Schulz - Multi-year Microsoft MVP Cloud and Data Center Management, former ten-time VMware vExpert. Author of Data Infrastructure Insights (CRC Press), Software Defined Data Infrastructure Essentials (CRC Press), as well as Cloud and Virtual Data Storage Networking (CRC Press), The Green and Virtual Data Center (CRC Press), Resilient Storage Networks (Elsevier) and twitter @storageio. Also visit www.picturesoverstillwater.com to view various UAS/UAV e.g. drone based aerial content created by Greg Schulz. Courteous comments are welcome for consideration. First published on https://storageioblog.com any reproduction in whole, in part, with changes to content, without source attribution under title or without permission is forbidden.
All Comments, (C) and (TM) belong to their owners/posters, Other content (C) Copyright 2006-2021 Server StorageIO and UnlimitedIO. Visit our companion site https://picturesoverstillwater.com to view drone based aerial photography and video related topics. All Rights Reserved. StorageIO is a registered Trade Mark (TM) of Server StorageIO.
2019 Happy Holidays Seasons Greetings, here's a video from our companion site Pictures Over Stillwater with holiday lights and Stillwater Lights.
Learn and view more via the following links:
Additional learning experiences along with common questions (and answers), as well as tips can be found in Software Defined Data Infrastructure Essentials book.
2019 Happy Holidays Seasons Greetings.
Ok, nuff said, for now.
Cheers GS
Greg Schulz - Multi-year Microsoft MVP Cloud and Data Center Management, ten-time VMware vExpert. Author of Data Infrastructure Insights (CRC Press), Software Defined Data Infrastructure Essentials (CRC Press), as well as Cloud and Virtual Data Storage Networking (CRC Press), The Green and Virtual Data Center (CRC Press), Resilient Storage Networks (Elsevier) and twitter @storageio. Also visit www.picturesoverstillwater.com to view various UAS/UAV e.g. drone based aerial content created by Greg Schulz. Courteous comments are welcome for consideration. First published on https://storageioblog.com any reproduction in whole, in part, with changes to content, without source attribution under title or without permission is forbidden.
All Comments, (C) and (TM) belong to their owners/posters, Other content (C) Copyright 2006-2021 Server StorageIO and UnlimitedIO. All Rights Reserved. StorageIO is a registered Trade Mark (TM) of Server StorageIO.
Join me for a free webinar Cloud Ready Data Protection for Hybrid Data Centers and Data Infrastructures 11AM PT Thursday July 11th produced by Redmond Magazine sponsored by Quest Software.
Hybrid cloud and on-prem data centers are in your future if not already a reality. In addition to using public cloud and on-prem resources, your environment is likely a mix of many different operating systems, applications and servers (virtual and physical), along with multiple backup and recovery technologies.
In this engaging, interactive webinar, we will look at trends, issues, and challenges, as well as provide best practices in what you can do to address them today. You’ll learn how to simplify and streamline your system, application and data protection in both the cloud and data center without compromise, all while removing complexity and cost.
Join Microsoft MVP, VMware vExpert and IT analyst Greg Schulz of Server StorageIO along with Michael Gogos, Data Protection expert from Quest, as they discuss how to:
Learn more about world backup day, recovery and data protection along with other related topics via the following links:
Additional learning experiences along with common questions (and answers), as well as tips can be found in Software Defined Data Infrastructure Essentials book.
I look forward to you joining Michael Gogos of Quest Software and myself on Thursday July 11th 11AM PT for our interactive discussion (bring your questions) around Cloud Ready Data Protection for Hybrid Data Centers and what you can do today (Register here).
Ok, nuff said, for now.
Cheers GS
Greg Schulz - Multi-year Microsoft MVP Cloud and Data Center Management, ten-time VMware vExpert. Author of Data Infrastructure Insights (CRC Press), Software Defined Data Infrastructure Essentials (CRC Press), as well as Cloud and Virtual Data Storage Networking (CRC Press), The Green and Virtual Data Center (CRC Press), Resilient Storage Networks (Elsevier) and twitter @storageio. Also visit www.picturesoverstillwater.com to view various UAS/UAV e.g. drone based aerial content created by Greg Schulz. Courteous comments are welcome for consideration. First published on https://storageioblog.com any reproduction in whole, in part, with changes to content, without source attribution under title or without permission is forbidden.
All Comments, (C) and (TM) belong to their owners/posters, Other content (C) Copyright 2006-2021 Server StorageIO and UnlimitedIO. All Rights Reserved. StorageIO is a registered Trade Mark (TM) of Server StorageIO.
Posted on July 23, 2018 at 23:52 0 Comments 0 Likes
Here is the 2018 Hot Popular New Trending Data Infrastructure Vendors To Watch which includes startups as well as established vendors doing new things. This piece follows last year’s hot favorite trending data infrastructure vendors to watch…
ContinuePosted on May 10, 2018 at 17:32 0 Comments 0 Likes
This is part one of a five-part series about Dell Technology World 2018 announcement summary. Last week (April 30-May 3) I traveled to Las Vegas Nevada (LAS) to attend Dell Technology World 2018 (e.g., DTW 2018) as a guest of Dell (that is a disclosure btw). There were several announcements along with…
Posted on April 6, 2018 at 20:07 0 Comments 0 Likes
Have you heard about the new CLOUD Act data regulation?
The new CLOUD Act data regulation became law as part of the recent $1.3 Trillion (USD) omnibus U.S. government budget spending bill passed by Congress on March 23, 2018 and signed by…
ContinuePosted on April 2, 2018 at 23:54 0 Comments 0 Likes
Microsoft Windows Server 2019 Insiders Preview has been announced. Windows Server 2019 in the past might have been named 2016 R2 also known as a Long-Term Servicing Channel (LTSC) release. Microsoft recommends LTSC…
ContinuePosted on April 2, 2018 at 23:38 0 Comments 0 Likes
It's time for Data Protection Recovery Life Post World Backup Day Pre GDPR Start Date.
The annual March 31 world backup day focus has come and gone once again.
However, that does not mean data protection including…
ContinueWelcome to
The Data Center Professionals Network
Connecting data center industry professionals worldwide. Free membership for eligible professionals.
© 2023 Created by DCPNet Admin.
Powered by
Comment Wall (2 Comments)
You need to be a member of The Data Center Professionals Network to add comments!
Join The Data Center Professionals Network
Hi Greg I am rizwan.Shall we become friends?